Project sheet
Sovereign high-availability cloud infrastructure
- Constraint
- Data subject to UMOA residency directives; 99.9% availability required.
- Decision
- Bare-metal 3-node K3s cluster in a local datacenter, run with GitOps (ArgoCD), with synchronous PostgreSQL replication.
- Outcome
- RPO = 0, no manual change in production, full rebuild from Git in under 30 minutes.
- Domain
- Banking & financeTelecoms
- Role
- Infrastructure Architect & DevSecOps
- Context
- Two projects for a regional financial institution and a major telecom operator in Benin: a bare-metal Kubernetes platform run with GitOps in a local datacenter, compliant with UMOA data-residency directives, with a 99.9% availability requirement.
- Experience
- KPS Groupe Bénin (IT services), Digital & Performance
- 3bare-metal K3s nodes, high availability
- RPO = 0synchronous PostgreSQL replication
- 99.9%availability required
- < 30 minfull rebuild from Git
What was done
- Bare-metal 3-node high-availability K3s cluster, run with GitOps (ArgoCD); provisioned with Terraform and Ansible.
- Data high availability: synchronous PostgreSQL replication (RPO = 0), Longhorn distributed storage; formalised PCA/PRA.
- Secured network: load balancer, network policies, site-to-site VPN, firewalls, automated TLS (cert-manager).
- Secrets management (Vault, External Secrets), strict RBAC, complete audit trail, in line with financial-sector requirements.
- No manual change in production: changes are versioned, reviewed and applied automatically; full infrastructure rebuild from Git in under 30 minutes.
- Linux (Ubuntu) administration of the physical servers, monitoring (Prometheus, Grafana, Loki) and production incident handling.
- Contribution to setting up the GitOps / CI-CD pipelines, monitoring and DevSecOps practices.
Technical environment
Select a technology to find other projects that use it.
Reading by layer
Choose a layer: it lights up in the diagram and its elements are described.
- Bare-metal 3-node K3s cluster, run with GitOps (ArgoCD); provisioned with Terraform and Ansible.
- No manual change in production: changes are versioned, reviewed and applied automatically.
- Full infrastructure rebuild from Git in under 30 minutes.
Simplified diagram built from the elements described in the sheet. No client data.